lipika-dewangan/cobe

COBE

Libraries:

python==3.8.10

transformers==4.1.1

pytorch==1.10.0

Run:

You can run the code for the task of cross-domain sentiment analysis for FDU-MTL dataset:

python run_COBE.py

You can change the parameters of --train_domains and --test_domains for different tasks.

As for cross-domain Amazon dataset, you need to change SentProcessor() to SentProcessor2().

After running, a result file can be found in the output directory (named test_results.txt).

Cite:

@article{luo2022mere,

title={Mere Contrastive Learning for Cross-Domain Sentiment Analysis},

author={Luo, Yun and Guo, Fang and Liu, Zihan and Zhang, Yue},

journal={arXiv preprint arXiv:2208.08678},

year={2022}

}